Re: Fast400 Value to iSeries community is less than zero

[Evan Harris <spanner@xxxxxxxxxx>]

James,

<Various snips>
>On 8 Nov 2001 thomas@inorbit.com wrote:
>
> > Understood, agreed, all well and good, etc., etc. But that still evades
> > the question that was asked which was what's the difference to the
> > customer between a software and a hardware limitation? Neither cost nor
> > effect would be different (assuming precise engineering of course).  The
> > limitation would exist either way. So, ??? Does it actually matter what
> > the mechanism is?
>
>Yes.  The small customer says, "Slower, smaller capacity hardware is
>cheaper to produce.  Those lower prices should mean I pay less for the
>machine.  I am getting screwed if I have to pay for high-powered hardware
>that is then artificially slowed."

In terms of the cost of producing a chip they would be wrong. It is cheaper
to produce one model of chip and strip/gear it down to fit the various
prize points the manufacturer determines the market will bear.


>The big customer says, "I paid a premium for high-performance.  My
>expensive machine has the same hardware capacity as the inexpensive
>machine.  Why don't I get a lower price if the same hardware can be sold
>more cheaply?  I am getting screwed!"

See argument above.

It's a fair argument that the way IBM has presented the CFINT governor is
not great, but my understanding of manufacturing this stuff is that if they
produced a machine for each processing tier that had that processing
capacity - and no more then - the smaller machines would be even MORE
expensive. Hence the use of an artificial method of creating the different
performance tiers.

Once upon a time when we hit the wall capacity wise the machine died and
you had no argument with IBM (except the cost to get to the next model :)
now we argue because we get a CFINT message about our interactive capacity
being exceeded.

You could also make a case that IBM erred by separating batch and
interactive - maybe to encourage us towards the type of server (read
"batch") processing that would reduce the need for green screen apps.

The reality is we just bitched because we couldn't get the batch processing
without doing any work.

OK maybe I've overstated the case in the last few paragraphs, but the
manufacturing costs incurred in creating high capacity chips do create a
real dilemma for IBM in how to deliver low end machines

Just my humble 2 cents worth :)

Regards
Evan Harris